Accurate Pelvic Fracture Detection for X-Ray and CT Images

1. An automated decision-support system for analysis of bone X-ray and computed tomography (CT) images to provide diagnostic and therapeutic recommendations, comprising: means for inputting X-ray bone images; first pre-processing means for automated segmenting of multiple bone X-ray image structures, said automated segmenting being performed on each structure individually using Active Shape Model (ASM) technique, a previously detected bone structure being used to begin detection of a next bone structure; fracture detection means receiving segmented multiple bone X-ray images from said first pre-processing means and outputting data identifying one or more detected fractures in an X-ray bone image; displacement calculation means receiving segmented multiple bone X-ray images from said pre-processing means and measuring displacements of bone segments; means for inputting CT bone images; second pre-processing means for automated segmenting of multiple bone CT image structures, template/shape matching means receiving segmented multiple bone CT images from said second pre-processing means and identifying segmented bone images; a processed trauma database receiving and storing data from said fracture detection means, said displacement calculation means, and said template/shape matching means; data processor means for accessing data in said processed trauma database, applying machine learning to accessed data, and generating rules for predicting injury severity; display means coupled with user interface means for receiving and displaying diagnostic information including images of detected bone fractures and predictions of injury severity from said data processor means, said user interface configured to receive user input of therapy to be administered to a patient; and a patient record database recording the diagnostic information and therapy to be administered for the patient.

2. The automated decision-support system according to claim 1 , wherein the ASM technique of said first pre-processing means incorporates cubic spline interpolation to regulate shape curvature of segmented X-ray bone images.

3. The automated decision-support system according to claim 2 , wherein the ASM technique of said first pre-processing means uses landmark positions to calculate symmetry and horizontal/vertical gap measures of segmented X-ray bone images.

4. The automated decision-support system according to claim 1 , wherein fracture detection means detects possible fractures in X-ray bone images using wavelet transform and edge tracing techniques.

5. The automated decision-support system according to claim 1 , wherein said second pre-processing means uses bone masks and seed growing techniques to extract objects potentially representing bone in a CT slice.

6. The automated decision-support system according to claim 5 , wherein said second pre-processing means uses bone masks in combination with shape matching to identify candidate bone objects.

7. The automated decision-support system according to claim 5 , wherein said second pre-processing means uses automated seeding of a Snake Model and uses the Snake Model to improve segmentation of bones.

8. The automated decision-support system according to claim 1 , further including means for inputting other information, including demographics and injury details, to said processed trauma database, said data processing means accessing said other information for generating rules for predicting injury severity and recommended therapy.

9. The automated decision-support system according to claim 8 , wherein the means for inputting other information inputs prior patient datasets to train a predictive model using machine learning by said data processor means, said data processor means using the predictive model to make predictions for a new patient.

10. The automated decision-support system according to claim 1 , wherein the X-ray bone images and the CT bone images are of a patient's pelvic region.

11. The automated decision-support system according to claim 1 , wherein the display means displays an X-ray image with highlighted potential fractures or displacements, a CT slice image with segmented bones showing fracture and areas of concern, patient data, and predicted patient outcome and recommendations for treatment.

12. The automated decision-support system according to claim 11 , wherein the display means further displays rules used to predict patient outcome with an option to view a rule tree, the display means displaying a rule tree when the option to do so is selected via the user interface means.

13. An automated decision-support method performed by a computer for analysis of bone X-ray and computed tomography (CT) images to provide diagnostic and therapeutic recommendations, comprising the steps of: inputting X-ray bone images; automated segmenting of multiple bone X-ray image structures, said automated segmenting being performed on each structure individually using Active Shape Model (ASM) technique, a previously detected bone structure being used to begin detection of a next bone structure; receiving segmented multiple bone X-ray images and outputting data identifying one or more detected fractures in an X-ray bone image; calculating displacement of received segmented multiple bone X-ray images from and measuring displacements of bone segments; inputting CT bone images; automated segmenting of multiple bone CT image structures, receiving segmented multiple bone CT images and identifying segmented bone images using template/shape matching; receiving and storing data of detected fractures, calculated displacement, and identified segmented bone CT images in a processed trauma database; accessing data in said processed trauma database, applying machine learning to accessed data, and generating rules for predicting injury severity; receiving and displaying diagnostic information including images of detected bone fractures and predictions of injury severity; receiving user input of therapy to be administered to a patient; and storing diagnostic information and therapy to be administered for the patient in a patient record database.

14. The automated decision-support method according to claim 13 , wherein the ASM technique incorporates cubic spline interpolation to regulate shape curvature of segmented X-ray bone images.

15. The automated decision-support method according to claim 14 , wherein the ASM technique uses landmark positions to calculate symmetry and horizontal/vertical gap measures of segmented X-ray bone images.

16. The automated decision-support method according to claim 13 , wherein detection of possible fractures in X-ray bone images is performed using wavelet transform and edge tracing techniques.

17. The automated decision-support method according to claim 13 , wherein bone masks and seed growing techniques are used to extract objects potentially representing bone in a CT slice.

18. The automated decision-support method according to claim 17 , wherein bone masks in combination with shape matching is used to identify candidate bone objects.

19. The automated decision-support method according to claim 17 wherein automated seeding of a Snake Model is performed and the Snake Model is used to improve segmentation of bones.

20. The automated decision-support method according to claim 13 , further including the step of inputting other information, including demographics and injury details, to said processed trauma database, said other information being accessed for generating rules for predicting injury severity and recommended therapy.

21. The automated decision-support method according to claim 20 , wherein the step of inputting other information inputs prior patient datasets to train a predictive model using machine learning by computer, said computer using the predictive model to make predictions for a new patient.

22. The automated decision-support method according to claim 13 , wherein the X-ray bone images and the CT bone images are of a patient's pelvic region.

23. The automated decision-support method according to claim 13 , wherein the step of receiving and displaying includes displaying an X-ray image with highlighted potential fractures or displacements, a CT slice image with segmented bones showing fracture and areas of concern, patient data, and predicted patient outcome and recommendations for treatment.

24. The automated decision-support method according to claim 23 , wherein the step of receiving and displaying further includes displaying rules used to predict patient outcome with an option to view a rule tree, and displaying a rule tree when the option to do so is selected via a user interface.

25. An automated decision-support system for analysis of bone X-ray images to provide diagnostic and therapeutic recommendations, comprising: means for inputting X-ray bone images; pre-processing means for automated segmenting of multiple bone X-ray image structures, said automated segmenting being performed on each structure individually using Active Shape Model (ASM) technique, a previously detected bone structure being used to begin detection of a next bone structure; fracture detection means receiving segmented multiple bone X-ray images from said first pre-processing means and outputting data identifying one or more detected fractures in an X-ray bone image; displacement calculation means receiving segmented multiple bone X-ray images from said pre-processing means and measuring displacements of bone segments; means for inputting other information, including demographics and injury details, to said processed trauma database, said data processing means accessing said other information for generating rules for predicting injury severity and recommended therapy; a processed trauma database receiving and storing data from said fracture detection means and said displacement calculation means; data processor means for accessing data in said processed trauma database, applying machine learning to accessed data, and generating rules for predicting injury severity, said means for inputting other information receiving prior patient datasets to train a predictive model using machine learning by said data processor means, said data processor means using the predictive model to make predictions for a new patient; display means coupled with user interface means for receiving and displaying diagnostic information including images of detected bone fractures and predictions of injury severity from said data processor means, said user interface configured to receive user input of therapy to be administered to a patient; and a patient record database recording the diagnostic information and therapy to be administered for the patient.